Lung microenvironments harbor Mycobacterium tuberculosis phenotypes with distinct treatment responses

ABSTRACT Tuberculosis lung lesions are complex and harbor heterogeneous microenvironments that influence antibiotic effectiveness. Major strides have been made recently in understanding drug pharmacokinetics in pulmonary lesions, but the bacterial phenotypes that arise under these conditions and their contribution to drug tolerance are poorly understood. A pharmacodynamic marker called the RS ratio® quantifies ongoing rRNA synthesis based on the abundance of newly synthesized precursor rRNA relative to mature structural rRNA. Application of the RS ratio in the C3HeB/FeJ mouse model demonstrated that Mycobacterium tuberculosis populations residing in different tissue microenvironments are phenotypically distinct and respond differently to drug treatment with rifampin, isoniazid, or bedaquiline. This work provides a foundational basis required to address how anatomic and pathologic microenvironmental niches may contribute to long treatment duration and drug tolerance during the treatment of human tuberculosis.

T o address the ongoing global tuberculosis (TB) epidemic, there is a critical need for new shorter, more effective combination antibiotic regimens. Current standard treatments require 6 months for the treatment of drug-susceptible TB and may take years for certain drug-resistant forms of TB. The development of novel regimens that cure TB more rapidly will require a better understanding of how drugs and regimens affect treatment-refractory bacterial populations.
One reason that TB demands prolonged therapy is the complexity and heterogeneity of TB lung lesions (1)(2)(3). Lesional microenvironments influence antibiotic effectiveness in two distinct ways. First, the variable architecture and composition of lesions, including differences in vascularization, fibrosis, inflammatory and immune cell infiltration and activation, and caseum formation, affect drug penetration and retention, resulting in drug-and lesion-specific pharmacokinetic (PK) profiles (2,(4)(5)(6). Second, variable physiochemical properties, including levels of O 2 , CO 2 , H + , carbon and nitrogen sources, micronutrients, and host immune milieu, elicit differing metabolic and stress respon ses in Mycobacterium tuberculosis (Mtb), resulting in phenotypically distinct bacterial populations in lesions that tolerate drug exposure differently. For example, Mtb in rabbit caseum is largely nonreplicating and exhibits extreme drug tolerance to many drugs (3). A similar dependence of antibiotic effectiveness on lesional phenotypes has been observed in nonhuman primates and marmosets (reviewed in (2)). Antibiotic effectiveness within a lesion depends on both PK (i.e., drugs must reach the pathogen at therapeutically relevant concentrations) and on the Mtb phenotype present (i.e., drugs must have activity against the lesion-specific phenotype) (7). While important strides have recently been made in understanding the PK distribution of antituberculosis agents into pulmonary lesions in diverse animal models of TB (6,8,9), there remains limited knowledge of lesion Mtb phenotypes and how these distinct phenotypes may respond differently to drug exposure (10)(11)(12)(13)(14).
Drug effects have traditionally been assessed based on colony forming units (CFU) which enumerate the burden of bacteria capable of growing on agar. We recently described the RS ratio (15), an alternative pharmacodynamic (PD) marker that is conceptually distinct from measures of bacterial burden. The RS ratio quantifies a key bacterial cellular process: ongoing rRNA synthesis. rRNA synthesis is fundamentally linked with bacterial replication (16,17). In the absence of drug treatment, the RS ratio is a proxy for Mtb replication (15). We have previously shown that the potency of a regimen in suppressing the RS ratio is associated with shortening time to nonrelapsing cure in mice (15).
Here we used both CFU and the RS ratio to evaluate the effect of three clinically relevant antibiotics in tissue microenvironments of the C3HeB/FeJ (Kramnik) mouse. We demonstrate that, in the absence of treatment, the Mtb populations of caseum, airway, remaining lung, and spleen exhibit distinct phenotypes that have a spectrum of ongoing rRNA synthesis. The RS ratio reveals that drugs with diverse mechanisms of action have different effects on Mtb microenvironment-specific phenotypes. Collectively, this work demonstrates that advancing beyond the enumeration of bacterial burden to consider the cellular activity of unique Mtb phenotypes provides insight into drug effect in lesions and how microenvironmental niches contribute to the long treatment durations required to cure TB.

Pathology of C3HeB/FeJ mouse microenvironments
After 12 wk of infection, the C3HeB/FeJ mice had diverse and complex pathology that enabled the evaluation of four microenvironments analogous to lesions observed in human TB (2,18,19). Discrete well-circumscribed granulomas classified as Type I lesions (2,18,19), in which the central caseum is bounded by a peripheral rim of enlarged, vacuolated macrophages and an outer ring of fibrosis and compressed alveolar lung tissue, were surgically incised to extract caseum ( Fig. 1A and B). The Mtb population in caseum is predominantly extracellular (18,19).
The second microenvironment was remaining lung after removal of Type I lesions. Remaining lung was dominated by nonnecrotic cellular Type III lesions consisting of an admixture of macrophages and lymphocytes with few neutrophils ( Fig. 1C and D). The Mtb population of remaining lung is predominantly intracellular within macrophages (18,19).
The third microenvironment was the airway which was sampled via bronchoalveolar lavage (BAL). In contrast to uninfected mice in which airway lavage is dominated by alveolar macrophages and epithelial cells (Fig. 1G), lavage from untreated Mtb-infected mice revealed higher proportions of neutrophils and fewer macrophages (Fig. 1H).
The fourth microenvironment was the spleen. Although spleen contains culturable Mtb, the pathological response is limited, consisting of microscopic aggregates of macrophages localized within lymphoid white pulp compartments ( Fig. 1E and F).

CFU burden and rRNA synthesis prior to drug treatment in key microenviron mental niches
Prior to drug treatment, CFU per gram was highest in caseum with levels at least 2 log 10 CFU higher than in remaining lung or spleen (minimum P < 0.00001) ( Fig. 2A; Table S1). Because it could not be normalized to the gram, CFU in BAL was not directly comparable to tissues. CFU in remaining lung was lower than in caseum but 0.78 log 10 CFU higher than in spleen (P = 0.02). The RS ratio identified a range of Mtb rRNA synthesis rates from lowest in caseum, intermediate in remaining lung and spleen, and highest in airway, indicating distinct microenvironment-specific phenotypes (Fig. 2B). Differences between the RS ratio in each microenvironment were statistically significant (Table S1).
Higher CFU was not associated with higher RS ratios. For example, the microenviron ment with the highest Mtb burden (caseum) had the lowest RS ratio, consistent with a large Mtb population and low ongoing rRNA synthesis, suggestive of a slowed bacterial replication rate in caseum. The less concentrated Mtb population in remaining lung had an RS ratio sixfold higher than caseum. CFU was similar in the spleen and airway but airway had a significantly higher RS ratio (P < 0.00001), suggesting an increased rate of bacterial replication in the oxygen-rich airway. As an additional point of reference, we assayed the RS ratio in the caseum surrogate model (20) used to assess nonspecific caseum drug binding (21,22) or drug activity against Mtb phenotypes that arise after incubation in lipid-rich caseum surrogate (23). The median RS ratio values were 45 in the ex vivo caseum surrogate and 33 in caseum from C3HeB/FeJ mice, suggesting the Mtb phenotypes are in similar states of rRNA synthesis and replication.

Drug effect in caseum
In caseum, INH and BDQ had no effect on the RS ratio (P = 0.97 and 0.99 relative to control, respectively). By contrast, RIF10 and RIF30 reduced the RS ratio by 2.6-fold and 4.1-fold (P = 0.04 and 0.003 relative to control, respectively). RIF30 reduced the RS ratio significantly more than INH or BDQ (P = 0.0004 and 0.001, respectively). In caseum, CFU did not identify significant differences between drugs. All drugs and doses reduced CFU in caseum by approximately one log 10 relative to the untreated controls.

Drug effect in spleen
In the spleen, INH and RIF10 reduced the RS ratio by 5.6-fold and 9.2-fold, respectively, relative to control (minimum P < 0.00001). Both RIF30 and BDQ had a significantly greater effect, reducing the RS ratio at least by 20-fold relative to control (minimum P < 0.00001) and significantly more than INH (minimum P < 0.00001). In contrast to caseum where BDQ had no effect on the RS ratio, BDQ was as effective as RIF30 in the spleen (P = 0.1). CFU did identify significant differences between drugs in the spleen. Specifically, RIF30 or BDQ reduced CFU burdens in the spleen more than INH (minimum P < 0.00001) or RIF10 (minimum P = 0.00005).

Drug effect in remaining lung
In remaining lung, INH reduced the RS ratio by 3.5-fold relative to control (P < 0.00001). RIF10, RIF30, and BDQ had significantly greater effects, reducing the RS ratio by at least 15-fold relative to control (minimum P < 0.00001) and significantly more than INH (minimum P =< 0.00001). Similar to results in the spleen, BDQ was as effective as RIF30 based on the RS ratio in remaining lung (P = 0.12). CFU revealed significant differences between drugs in remaining lung. Specifically, RIF30 or BDQ reduced CFU burdens more than INH (minimum P < 0.00001) or RIF10 (minimum P = 0.002).

Drug effect in airway
In the airway, INH reduced the RS ratio by three-fold relative to control (P = 0.01). RIF10, RIF30, and BDQ had significantly greater effects, reducing the RS ratio at least by eight-fold relative to control (minimum P < 0.00001). RIF30 had the greatest effect in airway, reducing the RS ratio significantly more than RIF10 or BDQ (minimum P = 0.03). In airway, CFU did not identify significant differences between drugs. All drugs and doses reduced CFU by approximately two log 10 relative to untreated controls in airway.

Effects of individual drugs across different microenvironments
Using the RS ratio to evaluate individual drugs in different microenvironments revealed drug-specific changes ( Fig. 4; Table S3 and S4). We evaluated both the change in RS ratio relative to control and the absolute RS ratio value following treatment.

Effect of INH in caseum, spleen, remaining lung, and airway
As noted above, INH caused a minimal change in the RS ratio relative to control in caseum (Fig. 4A). By contrast, INH reduced the RS ratio by 5.6-fold in spleen, 3.5-fold in remaining lung, and 3.0-fold in airway, representing a significantly greater reduction in all microenvironments relative to caseum (minimum P = 0.01). Following treatment with INH, the absolute value of the RS ratio did not differ significantly between caseum and spleen (P = 0.3) or remaining lung (P = 0.5) ( Fig. 4B; Table S4). Following treatment, the RS ratio remained significantly higher in airway than in any other microenvironment (minimum P = 0.01). The log 10 reduction in CFU relative to control was statistically indistinguishable between microenvironments, with the exception that CFU declined marginally more in airway than in caseum (P = 0.04) (Fig. 4C).

Effect of RIF10 in caseum, spleen, remaining lung and airway
RIF10 reduced the RS ratio by 2.6-fold relative to control in caseum (Fig. 4D). By contrast, RIF10 reduced the RS ratio by 9.2-fold in spleen, 15.5-fold in remaining lung, and 7.7-fold in airway, representing a significantly greater reduction in all other microenvironments than in caseum (minimum P = 0.01). Following treatment with RIF10, the absolute value of the RS ratio was similar in caseum, spleen, and remaining lung but remained significantly higher in airway (minimum P < 0.00001) ( Fig. 4E; Table S4). RIF10 reduced CFU to statistically indistinguishable degrees in spleen, remaining lung, and airway (2.0 to 2.4 log 10 CFU) (Fig. 4F). RIF10 reduced CFU significantly less in caseum than in remaining lung (P = 0.002) or airway (P = 0.009).

Effect of RIF30 in caseum, spleen, remaining lung and airway
RIF30 reduced the RS ratio by 4.1-fold relative to control in caseum (Fig. 4G). By contrast, RIF30 reduced the RS ratio by 20.4-fold in spleen, 32.6-fold in remaining lung. and 41.7-fold in airway, representing a significantly greater reduction in all other microenvir onments than in caseum (minimum P = 0.006). Following treatment with RIF30, the absolute value of the RS ratio was similar in all microenvironments ( Fig. 4H; Table S4). RIF30 reduced CFU significantly less in caseum than in spleen (P = 0.04) or remaining lung (P = 0.0001). Evaluation of exposure response showed that RIF30 reduced the RS ratio significantly more than RIF10 in spleen, remaining lung, and airway (minimum P = 0.004). RIF30 also reduced CFU significantly more in spleen and remaining lung than in caseum (minimum P = 0.002). For caseum, there was no significant difference between RIF30 and RIF10 for either the RS ratio or CFU.

Effect of BDQ in caseum, spleen, remaining lung and airway
BDQ had minimal effect on the RS ratio in caseum (Fig. 3J). By contrast, BDQ reduced the RS ratio by 33.5-fold in spleen, 22.1-fold in remaining lung, and 10.6-fold in air way, representing a significantly greater reduction in these microenvironments than in caseum (minimum P = 0.0008). Following treatment with BDQ, the absolute value of the RS ratio remained significantly higher in caseum than in spleen (P < 0.00001) or remaining lung (P = 0.0003) (Fig. 3K). BDQ reduced CFU significantly less in caseum than in any other microenvironment (minimum P = 0.004) (Fig. 3J).

Drug distribution within microenvironments
To determine how standard potency metrics compare to concentrations achieved in each lung microenvironment and to help interpret the efficacy results, we applied drug quantitation by liquid chromatography with tandem mass spectrometry (LC/MS-MS) of samples collected by laser-capture microdissection (LCM) in thin lesion sections. Quantification of LCM samples after 17 doses demonstrated that RIF10 was present at concentrations well above the serum-shifted minimum inhibitory concentration (fMIC) ( Table S5) in all sampled tissue compartments including caseum. Trough sampling 24 h after dosing showed unchanged RIF concentrations in caseum (Fig. 5A). At day 17, INH was present at concentrations well above fMIC in all sampled tissue compartments including caseum 1 h after dosing but rapidly exited, falling below fMIC after 24 h (Fig.  5B). At day 17, BDQ concentrations were similar at peak and trough and displayed a gradient in which concentrations observed in inner caseum were 16-to 28-fold lower than in the outer caseum, which showed to be proportional to the distance from the edge of the granuloma (Fig. 5C, E and F). The BDQ gradient was much more pronounced with a single dose of BDQ, wherein found levels of BDQ ranged from 1,490 ng/g (peak) to 2,465 ng/g (trough) in outer caseum to below the limit of quantification in inner caseum (Fig. 5D).

DISCUSSION
Application of the RS ratio and CFU burden demonstrated that lung microenvironments of the C3HeB/FeJ mouse harbor distinct Mtb phenotypes that have markedly different degrees of ongoing rRNA synthesis. Microenvironmental phenotypes appeared to be a determinant of drug effectiveness. All drugs had the lowest activity against Mtb populations in caseum. Even when tissue drug concentrations were above the serumshifted MIC, certain drugs were more effective against certain microenvironmental Mtb populations than others. Our in vivo observations in C3HeB/FeJ mice support and extend results showing higher drug tolerance and differential drug activity against Mtb in explanted rabbit caseum ex vivo (3,24). These results highlight the importance of considering PK and PD jointly. Combining a PD readout of a fundamental cellular process with lesional PK data may help to identify drugs and combinations that optimally target Mtb populations in caseum, a critical obstacle to shorter TB treatments.
Our findings are consistent with the long-standing hypothesis that one reason TB requires prolonged therapy is that unique tissue microenvironments harbor "special populations" that vary in their ability to withstand drug exposure (1,7). The RS ratio showed that Mtb maintains low ongoing rRNA synthesis in caseum, indicating a quies cent slowly replicating population. This is consistent with the ex vivo rabbit caseum model in which CFU and chromosomal equivalents showed no evidence of Mtb replication (3). Indeed, our RS ratio results indicated a similarly low level of rRNA synthesis for Mtb in caseum in vivo compared to caseum surrogate ex vivo, which in the absence of drug treatments, is indicative of a dramatically reduced replication rate. Conversely, the Mtb populations of other environments appear to be replicating more quickly, with the highest RS ratios in the oxygen-rich airway.
Although caseum had a low RS ratio suggesting slow growth, caseum had a CFU burden 2.2 log 10 higher than remaining lung. It has previously been shown that C3HeB/FeJ mice experience a phase of rapid Mtb replication early in granuloma development (19,25) that is followed by slowing of replication as the lesion becomes hypoxic and necrotic (19). It is also possible that the absence of intact immune effector cells in caseum results in a decreased rate of immune elimination relative to remaining lung.
PD markers play a central role in drug evaluation because they are the readout used to assess drug effect. The historical standard PD marker is an enumeration of the burden of Mtb capable of growth on solid culture (i.e., CFU) (26,27). The RS ratio introduces a fundamentally different type of readout that measures a central bacterial cellular process rather than total CFU burden. The difference between RS ratio and CFU was highlighted in a recent in vitro analysis that showed a low correlation between drug effect on RS ratio and CFU (28). As highlighted below, this new readout of a physiologic property revealed differences in drug effects that were not discernible based on CFU, thereby providing a new perspective on the effect of individual drugs against different microenvironmental phenotypes.

Full-Length Text
Antimicrobial Agents and Chemotherapy INH illustrates that achieving a concentration that exceeds the fMIC is insufficient if a microenvironment harbors an Mtb phenotype that is capable of withstanding a drug's mechanism of action. For the quiescent Mtb population of caseum that likely has a low need for ongoing mycolic acid synthesis, effective concentrations of INH had no effect on the RS ratio. By contrast, for the actively replicating Mtb populations in spleen, remaining lung, and airway, interruption of new cell wall synthesis by INH reduced the RS ratio by 3.0-to 5.6-fold. This is consistent with a body of evidence showing that INH is effective primarily against replicating Mtb, with minimal activity against nonreplicating Mtb (3,29,30). In contrast to the RS ratio, CFU did not reveal differences between the effect of INH in different microenvironments, with the exception that CFU declined marginally more in airway than caseum. INH highlights the importance of bacterial phenotype and the information provided by a new molecular marker of rRNA synthesis.
RIF reduced the RS ratio in all sampled tissue compartments, indicating that RIF has activity against the phenotypes found in all microenvironments. Because there has long been uncertainty about optimal rifampin dosing (31, 32), we tested the humanequivalent standard dose (10 mg/kg) as well as a human-relevant higher dose (30 mg/ kg). Consistent with evidence that higher-dose rifampin has greater activity in preclini cal animals and humans (33), RIF10 and RIF30 demonstrated a strong dose-response relationship in spleen, remaining lung, and airway. Nonetheless, both RS ratio and CFU indicated that RIF was less effective in caseum than in any other microenvironment. This highlights that caseum phenotypes are better able to tolerate even high concentrations of RIF, the prototypical sterilizing agent that is the backbone of existing front-line TB treatment.
BDQ represents a different example. In microenvironments that achieved high BDQ concentrations (e.g., remaining lung), BDQ matched the potency of RIF30 as measured by RS ratio and CFU. By contrast, drug distribution of BDQ into caseum appeared to be proportional to the number of doses administered and inversely correlated with distance from the outer caseum edge. Spatial drug distribution studies by LCM LC/ MS-MS suggested that after 2.5 wk of treatment the concentration of BDQ ranged from 14.6 µg/g in outer caseum to as low as 0.25 µg/g in inner caseum and the effect on RS ratio was correspondingly negligible. Comparing the concentration gradient after 1 and 17 doses (Fig. 5C and D) suggests that BDQ slowly penetrates avascular caseum and may not have reached steady state at the 2.5 wk mark, consistent with the slow onset of efficacy in an early bactericidal activity Phase II trial (34). The current results do not indicate what the effect of BDQ might be were adequate caseum concentrations achieved.
Our observation that all drugs have diminished effectiveness in C3HeB/FeJ mouse caseum is broadly consistent with previous results from the rabbit ex vivo caseum model (3). Using CFU as a readout, Sarathy et.al. found that the caseum minimum bactericidal concentration of first-and second-line drugs was higher in rabbit caseum than under standard in vitro conditions with replicating cultures. Because the RS ratio is a direct molecular PD readout that does not require confounding steps such as recovery and outgrowth of bacteria, the RS ratio enabled this first comparison of phenotypes and drug effects across tissue microenvironments. The RS ratio extends and complements results from the ex vivo rabbit caseum model by measuring a fundamental cellular process rather than bacillary burden.
In contrast to our results, a recent manuscript that evaluated a fluorescent reporter Mtb strain in C3HeB/FeJ mouse lesions reported that Mtb was more actively replicating in the caseum core than in the cuff, and the Mtb population of the core showed greater susceptibility to INH (25). Importantly, the previous report evaluated a different timepoint than tested here. Lavin and Tan (25) evaluated mice 6 wk post-infection (a timepoint at which previous studies would suggest lesions are in the maturation process (19)) and identified rapidly replicating caseum phenotypes that were highly susceptible to INH. By contrast, we evaluated mice starting treatment at 10 wk post-infection (a timepoint at which lesions are mature and well encapsulated with high Mtb burden) and identified slowly replicating caseum phenotypes that were INH tolerant. In combination, these observations appear to highlight how the evolution of lesions and phenotypes over time dictates drug effectiveness. Time-course studies using a single consistent assay method are needed to confirm this hypothesis. This report has several limitations. This proof-of-concept study included limited PK measurements that were insufficient for statistical pharmacokinetic modeling. Future studies can maximize the value of lesional analysis by pairing lesional PD (based on RS ratio and CFU) with comprehensive lesional PK profiles. Second, the RS ratio provides a snapshot of average rRNA synthesis across an entire Mtb population. Using in situ hybridization with single-bacillary resolution, we have previously shown that caseum harbors multiple Mtb phenotypes with differing degrees of ongoing rRNA synthesis (8,15). In future studies, the joint PK/PD evaluation demonstrated here can be extended with even more spatially granular methods.
This study has shown that different tissue microenvironments harbor phenotypically distinct Mtb populations that respond to drug treatment differently. As a molecular readout of a fundamental cellular process, the RS ratio provides a new metric for evaluating efficacy in lesions that does not rely on bacillary burden. This report points to a new era in which lesional PK can be combined with molecular assays of bacterial cellular processes like the RS ratio to design regimens that optimally target caseum phenotypes and achieve the end goal of shorter, more effective TB treatments.

Drug susceptibility profiling
The minimum inhibitory concentration (MIC) was determined for RIF, INH, and BDQ against Mtb Erdman in 7H9 media supplemented with 0.2% [vol:vol] of glycerol and 10% [vol:vol] of albumin-dextrose-catalase (ADC), with 0.05% [vol:vol] of Tween-80 (7H9 media). MICs were determined by a broth microdilution assay using two-fold serial drug dilutions. The lowest consecutive antimicrobial concentration that showed a ≥80% reduction in OD 600 relative to drug-free control wells was regarded as the MIC. In parallel, the MICs were also determined in the presence of 4% (wt/vol) of human serum albumin (Sigma # A1653) which is defined here as the serum-shifted MIC.

Animals
Female-specific pathogen-free C3HeB/FeJ mice, aged 8-10 weeks, were purchased from Jackson Laboratories (Bar Harbor, ME). Mice were housed in an animal bio-safety level III (ABSL-3) facility employing autoclaved bedding, water, and mouse chow ad libitum. All procedures were approved by Colorado State University Institutional Animal Care and Use Committee (IACUC) (Reference numbers of approved protocol: KP 1515).

Aerosol infection
C3HeB/FeJ mice were exposed to a low-dose aerosol infection with the Mtb Erdman strain (TMCC 107) using a Glas-Col inhalation exposure system resulting in an average of 50-75 bacteria in lungs 1 d following aerosol (8,35). Infected mice were observed and weighed at least once a week. Starting from day 21 until the start of therapy, mice were observed and weighed two to three times per week, due to the increased incidence of morbidity and mortality associated with clinical TB disease. Any mice exhibiting clinical symptoms of illness were humanely euthanized.

Drug preparation and drug treatment
Rifampin (Sigma) and isoniazid (Sigma) were dissolved in sterile water. Bedaquiline fumarate (PharmaBlock) was dissolved in 20% (wt/vol) of 2-hydroxypropyl-β-cyclodex trin and formulated as described previously (36,37). For drug efficacy experiments, C3HeB/FeJ mice were randomized (n = 12 per group) and dosed once daily starting 10 wk post-aerosol with rifampin at 10 mg/kg or 30 mg/kg, isoniazid at 25 mg/kg, or bedaquiline fumarate at 25 mg/kg, 7 d per week by oral gavage in 0.2 mL of volume for the total number of doses indicated.

Histological analysis
A separate group of mice were sacrificed and whole lungs were fixed for histopathology purposes. Lung perfusions were performed by clamping the caudal vena cava with straight-tipped hemostats and cutting the vena cava between the hemostats and the liver to allow blood to drain. A 24-guage 0.75-inch catheter was inserted into the right ventricle of the heart and blood was flushed using phosphate-buffered saline (PBS) with 0.04% [wt:vol] of ethylenediaminetetraacetic acid followed by 10 mL of 4% paraformal dehyde (PFA). Lungs were recovered and placed into a histology cassette and incubated for 48 h in 4% of PFA for 48 h prior to transfer to PBS. Lungs were sectioned and stained with hematoxylin-eosin (H&E).

Collection of samples for bacterial enumeration and RS ratio
Each mouse was individually euthanized by CO 2 narcosis followed by cardiac puncture. Airway was sampled by BAL by passing a total of 2.5 mL of PBS into the lungs through a 24-guage 0.75-inch catheter inserted into the trachea. An amount of 0.5 mL of BAL was used for CFU enumeration, and an amount of 0.5 mL was placed into 2 mL of CAMM-RPS buffer to preserve RNA. The remaining BAL was collected by cytospin for Wright's Giemsa staining. Spleens were aseptically collected. Lungs were recovered, photographed, and diagramed. To collect inner caseum and avoid cross-contamination from the granuloma's cellular cuff, we first identified lung lobes featuring large encapsulated caseous necrotic granulomas (Type I). The granuloma capsule was carefully incised with a scalpel and the granuloma sides were gently squeezed with tweezers to open the incision and expose the inner caseum. A small spoon was inserted through the incision and used to scoop out a small amount (~0.01 to 0.03 g) of the inner caseum which was split to two cryovials. All tissue samples were divided in half and weights were recorded. Samples were flash-frozen in liquid nitrogen and stored at −80°C prior to further processing.
Tissues were disrupted with a tissue homogenizer (Precellys, Bertin Instruments, Rockville, MD) in PBS plus 10% [wt/vol] of bovine serum albumin for CFU enumeration on 7H11 agar further supplemented with 0.4% activated charcoal to prevent drug carry-over as described previously (15). Colonies were enumerated after at least 28 d of incubation at 37°C and the plates were incubated for 8 wk to ensure that all viable colonies were detected. For RNA extraction, tissues were homogenized in Trizol (15).

Caseum surrogate model
PMA-differentiated THP-1 macrophages were infected with irradiated Mtb (BEI Resour ces) at an approximate multiplicity of infection of 1:50. Foamy macrophages were washed three times with PBS, followed by three freeze-thaw cycles to lyse the cells and incubated at 75°C for 20 min to denature proteins in the matrix. The caseum surrogate matrix was rested for 3 d at 37°C and stored at −20°C prior to inoculation.

Collection of samples from caseum surrogate
Mtb HN878 grown to mid-log phase in 7H9 media was harvested by centrifugation, suspended in sterile water, and inoculated into the caseum surrogate (22,23) at 10 8 CFU/mL. The inoculated caseum surrogate was mixed briefly using 1.4 mm zirconium oxide beads (CK14 soft tissue; Precellys, Bertin Instruments, Rockville, MD) to achieve a homogeneous suspension. After 8-wk incubation in a sealed tube at 37°C to allow for physiologic and metabolic adaptation to the lipid-rich environment, CAMM-RPS buffer was added at 3 × the total caseum surrogate volume to preserve RNA.

Collection of samples for pharmacokinetic analysis
For the PK experiments, 9-10 C3HeB/FeJ mice were dosed starting 10 wk post-aerosol as described above. Treatment occurred 7 d per week for up to a total of 17 doses. Drugs were prepared as described above for the efficacy studies. Mice were euthanized, and plasma and tissues were collected at two timepoints, selected based on the plasma C max (1 or 5 h) and C min (24 h). Whole blood was obtained via cardiac puncture and processed in plasma separator tubes (Becton, Dickinson and Co., Franklin Lakes, NJ, USA) and centrifuged at 3,000 × relative centrifugal force for 2 min at 4°C, aliquoted into Eppendorf microcentrifuge tubes and stored at −80°C until analysis. Mice with pronounced lung pathology were selected to collect samples for spatial drug quantitation by gravity-assis ted LCM. Briefly, whole lung samples were collected on clear disposable base molds (Fisher Scientific, Hampton, NH, USA). Using forceps, tissues were collected on trays placed on a prechilled aluminum block held in liquid nitrogen, and frozen within 1-2 min. Tissue trays containing frozen lobes were wrapped in foil squares, placed individu ally into labeled zip-lock bags, and immediately transferred onto dry ice.

Drug quantitation by HPLC coupled to tandem mass spectrometry (LC/MS-MS)
Drug levels in tissue were determined by spatial quantitation in thin tissue sections by LCM followed by LC/MS-MS analysis of microdissected areas (38). The benefit of the LCM approach is the ability to obtain absolute drug levels in defined areas of Type I lung lesions such as caseum (the core of the caseous necrotic lesion) without crosscontamination. The tissue sections of 25 µm thickness were cut from infected mouse lung biopsies using a Leica CM 1860UV (Buffalo Grove, IL) and thaw-mounted onto Leica PET-Membrane FrameSlides (Buffalo Grove, IL) of 1.4 µm thickness. Approximately 10 µm thick tissue sections were thaw-mounted onto standard glass microscopy slides for H&E staining. Sequential rings of necrotic tissue proceeding from the border with the cellular rim to the core of the caseous compartment were demarcated for each tissue section by creating a mask of caseum and processing it using the Exact Euclidean Distance Transform plugin in ImageJ (National Institutes of Health, MD). Sequential rings of necrotic tissue were then dissected using a Leica LMD6 system (Buffalo Grove, IL). Dissected lesion tissues were collected into standard PCR tubes of 0.25 mL and immediately transferred to −80°C.
RIF, BDQ, INH, and verapamil (VER) were purchased from Sigma Aldrich (St. Louis, MO, USA); RIF-d8, BDQ-d6, and INH-d4 were purchased from Toronto Research Chemi cals (Ontario). Drug-free lung and K 2 EDTA plasma from CD-1 mice was obtained from BioIVT (Westbury, NY) for use as blank matrices to build standard curves. Neat 1 mg/mL dimethyl sulfoxide stocks were serially diluted in 50:50 acetonitrile/water to create a standard curve and quality control spiking solutions and 10 µL of neat spiking solutions were added to 2 µL of lesion homogenate prior to extraction. On the day of analysis, samples were extracted by adding 50 µL of extraction solution (acetonitrile/methanol, 1:1) containing 1, 10, 4, and 5 ng/mL of VER, INH-d4, RIF-d8, and BDQ-d6, respectively. Extracts were sonicated for 5 min and centrifuged at 10,000 rpm for 5 min, and 40 µL of supernatant was transferred to 96-well deep plates for LC-MS/MS analysis. For INH quantitation, 40 µL of 2% cinnamaldehyde in methanol was added to derivatize isoniazid prior to analysis. For RIF quantitation, 5 µl of 75 mg/mL ascorbic acid was added to extracts to stabilize RIF during analysis.
LC/MS-MS analysis was performed on a Sciex Qtrap 6500 + triple-quadrupole mass spectrometer coupled to a Shimadzu Nexera X2 UHPLC system. Chromatography was performed on an Agilent Zorbax SB-C8 column (2.1 × 30 mm; particle size 3.5 µm) using a reverse phase gradient. Deionized water with 0.1% formic acid (FA) was used for the aqueous mobile phase and 0.1% FA in acetonitrile for the organic mobile phase. Multiple-reaction monitoring (MRM) of precursor/fragment transitions in the electro spray positive-ionization mode was used to quantify the analytes. MRM transitions of 823. 30

RNA extraction and RS ratio profiling
RNA was extracted from tissue and BAL samples after homogenization in Trizol (15). Briefly, eukaryotic cellular debris was pelleted and the lysate was mixed with chloro form. A solution of 50% isopropanol, 0.8M sodium citrate, and 1.2M sodium chloride was mixed with the aqueous phase and nucleic acids were precipitated overnight at 4°C. Nucleic acids were pelleted by centrifugation, washed twice with 70% ethanol, and resuspended in water. DNA was digested in a multi-step process, with the initial digestion performed using two additions of Promega RQ1 DNase followed by a 15min incubation at 37°C after each DNase addition. RNA was further purified and DNA was digested on a Maxwell RSC instrument (Promega) with the Maxwell RSC simplyRNA tissue kit. Purification was performed following the manufacturer's instructions, with DNase added at twice the recommended amount.
Purified RNA was reverse transcribed to cDNA using SuperScript III VILO cDNA synthesis kit (Invitrogen) at 42°C for 120 min. The RS ratio was measured by quantification of ETS1 and 23S rRNA gene transcripts in the cDNA using droplet digital PCR in a duplexed reaction. Droplet digital PCRs were performed using the QX200 Droplet Digital PCR system with ddPCR SuperMix for Probes (no dUTP) (Bio-Rad). The RS ratio was calculated from each duplexed reaction by the QX200 Droplet Digital PCR system software (QuantaSoft AP, Bio-Rad).

Statistical analysis
A one-way ANOVA was performed to compare the effect of different drugs or microen vironments on CFU and the RS ratio followed by a multiple comparison analysis of variance using a one-way Tukey test. Differences were considered significant at the 95% confidence level (P < 0.05). Data management, plotting, and post-modeling analysis were conducted using R (v 4.2.1; R Development Core Team, Vienna, Austria).